Prune bias from the root: Bias removal and fairness estimation by pruning sensitive attributes in pre-trained DNN models
Bibliographic record
Abstract
Context: Deep learning models (DNNs) are widely used in high-stakes decision-making domains, but they often inherit and amplify biases present in training data, leading to unfair predictions. Given this context, fairness estimation metrics and bias removal methods are required to select and enhance fair models. However, we found that existing metrics lack robustness in estimating multi-attribute group fairness. On the other hand, existing post-processing bias removal methods often focus on group fairness and fail to address individual fairness or optimize along multiple sensitive attributes. Objectives: In this study, we explore the effectiveness of attribute pruning in both bias removal and multi-attribute group fairness estimation. Methods: Attribute pruning is a simple yet effective post-processing technique that enforces individual fairness by zeroing out sensitive attribute weights in a pre-trained DNN’s input layer. To ensure the generalizability of our results, we conducted experiments on 25 models and 3 widely used datasets, and compared attribute pruning's performance with 3 baseline post-processing methods. Results: According to the results, we demonstrate that single-attribute pruning can stably improve group fairness while preserving model accuracy, outperforming conventional post-processing methods in 18 out of 25 cases. Additionally, we explore multi-attribute pruning, showing its effectiveness in mitigating bias across multiple sensitive attributes while maintaining competitive accuracy. Furthermore, we introduce a novel multi-attribute fairness estimation approach based on attribute pruning, enabling fairness-aware model selection through subgroup-wise true positive rate analysis. Conclusion: Our findings highlight the potential of attribute pruning as an efficient, interpretable, and generalizable approach for bias mitigation and estimation in DNNs. We advocate for its adoption as a comparison baseline in fairness-aware AI development and encourage further exploration of its applications in fairness estimation.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.013 | 0.038 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.002 | 0.001 |
| Scholarly communication | 0.002 | 0.003 |
| Open science | 0.002 | 0.003 |
| Research integrity | 0.003 | 0.007 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".